Translational Psychiatry www.nature.com/tp ARTICLE OPEN Genetic association study of childhood aggression across raters, instruments, and age © The Author(s) 2021 Childhood aggressive behavior (AGG) has a substantial heritability of around 50%. Here we present a genome-wide association meta- analysis (GWAMA) of childhood AGG, in which all phenotype measures across childhood ages from multiple assessors were included. We analyzed phenotype assessments for a total of 328 935 observations from 87 485 children aged between 1.5 and 18 years, while accounting for sample overlap. We also meta-analyzed within subsets of the data, i.e., within rater, instrument and age. SNP-heritability for the overall meta-analysis (AGGoverall) was 3.31% (SE = 0.0038). We found no genome-wide significant SNPs for AGGoverall. The gene- based analysis returned three significant genes: ST3GAL3 (P = 1.6E–06), PCDH7 (P = 2.0E–06), and IPO13 (P = 2.5E–06). All three genes have previously been associated with educational traits. Polygenic scores based on our GWAMA significantly predicted aggression in a holdout sample of children (variance explained = 0.44%) and in retrospectively assessed childhood aggression (variance explained = 0.20%). Genetic correlations (rg) among rater-specific assessment of AGG ranged from rg = 0.46 between self- and teacher-assessment to rg = 0.81 between mother- and teacher-assessment. We obtained moderate-to-strong rgs with selected phenotypes from multiple domains, but hardly with any of the classical biomarkers thought  to be associated with AGG. Significant genetic correlations were observed with most psychiatric and psychological traits (range rg : 0.19–1.00), except for obsessive-compulsive disorder. Aggression had a negative genetic correlation (rg= ~−0.5) with cognitive traits and age at first birth. Aggression was strongly genetically correlated with smoking phenotypes (range rg : 0.46–0.60). The genetic correlations between aggression and psychiatric disorders were weaker for teacher-reported AGG than for mother- and self-reported AGG. The current GWAMA of childhood aggression provides a powerful tool to interrogate the rater-specific genetic etiology of AGG. Translational Psychiatry (2021)11:413 ; https://doi.org/10.1038/s41398-021-01480-x INTRODUCTION well-timed coercion strategy are generally more liked by their There is a variety of phenotypic definitions of aggressive peers and score higher on social dominance [22]. behavior (AGG), from broadly defined externalizing problems Despite a heritability of roughly 50% [5, 23], genome-wide to narrow definitions like chronic physical aggression [1]. association studies (GWASs) on childhood AGG have not identified Generally any action performed with the intention to harm genome-wide significant loci that replicated [1]. Childhood another organism can be viewed as AGG [2, 3]. AGG is cohorts often have rich longitudinal data and assessments from considered a common human behavior [4], with people varying multiple informants and we aimed to increase power to detect in the degree of AGG they exhibit [5]. Children typically display genomic loci via multivariate genome-wide association meta- AGG early in life, after which symptoms tend to diminish [6, 7], analysis (GWAMA) across genetically correlated traits [24, 25]. In although in some individuals AGG persists into adulthood [8]. AGG, twin studies have reported moderate to high genetic AGG is also part of numerous childhood and adult disorders [9], correlations among instruments, raters, and age [26–29]. Child- including oppositional defiant disorder and conduct disorder hood behavior can be context dependent, with teachers, fathers, (CD) [10]. In its extreme forms, AGG may be considered a and mothers each observing and rating aggression against a disorder by itself—inflicting a huge personal and financial different background. Teachers are typically unrelated to the child, burden on the individual, their relatives, friends, and society as and see the child in the context of a structured classroom and can a whole [11]. In general population studies, AGG is commonly judge the child’s behavior against that of other pupils. Parents treated as a quantitative trait, and pathological AGG has been share part of their genome with their offspring and, most often, a argued to be best seen as the extreme end of such a continuum household. Parental genomes also influence the home environ- [12–14]. Childhood AGG co-occurs with many other behavioral, ment, and it is predominantly within this context that parents emotional, and social problems [15, 16] and is associated with observe the child’s behavior. Multiple assessments of aggression increased risk of developing negative outcomes later in life, by teachers, fathers, and mothers, by different instruments and at including cannabis abuse [17], criminal convictions [18], anxiety different ages, provide information that may be unique to a disorder [19], or antisocial personality disorder [20]. Not all specific context and therefore may capture context-dependent associated outcomes are harmful [21]. For example, children who expression of AGG. These considerations support an approach in learn to control their impulses and apply aggressive acts as a which all AGG data are simultaneously analyzed, while retaining A full list of authors and their affiliations appears at the end of the paper. Received: 24 May 2020 Revised: 11 April 2021 Accepted: 20 May 2021 H.F. Ip et al. 2 Table 1. (a) Multivariate test statistic in the meta-analysis of results based on overlapping samples; (b) expected value for the cross-trait-intercept; (c) effective sample size for a GWAMA. PP w Z Zmulti;j ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i¼1 ji ji ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi test statistic for jth SNP. P is the number of GWASs across that we run the meta-analysis; (a) Multivariate PP PP PP pffiffiffiffiffiffiffiffiffi wji Vji þ i¼1 i¼1 k¼1 wji wjk CTIik for i≠k wji ¼ Nji h2SNP;i is the weight given to the jth SNP in GWAS i, with h2SNP;i being the SNP-heritability of the trait analyzed in GWAS i; and Vji ¼ 1 represents the variance of the distribution of Zji under the null hypothesis of no effect. Nr CTIik ¼ pffiffiffiffiffiffiffiffi s p (b) Cross-trait intercept between GWAS i and k. Ns represents the sample overlap; rp indicates the Nji Njk phenotypic correlation; Nji and Njk are the sample sizes at SNP j for, respectively, GWASs i and k pffiffiffiffiffiffi pffiffiffiffi Neff ¼ NT CTI1 N (c) N is an P-sized vector of sample sizes, and CTI is the P × P matrix of cross-trait-intercepts. the ability to analyze the data by rater. Our analyses include correlations among cohorts were set to zero and omitted from repeated observations on the same subject, which requires Supplementary Fig. 1. Phenotypic correlations of zero also appropriate modeling of the clustered data, since the covariance correspond to independent samples within a cohort. For GWASs between test statistics becomes a function of a true shared with sample overlap, most phenotypic correlations ranged genetic signal and the phenotypic correlation among outcomes between 0.1 and 0.4, with a median value of 0.29. When [29]. We developed an approach that allowed inclusion of all stratified by rater, phenotypic correlations were more heavily measures for a child—e.g., from multiple raters at multiple ages— centered around 0.4 (see Supplementary Fig. 1). The maximum and resolved issues of sample overlap at the level of the meta- number of correlations within cohort at a specific age is three analysis. By doing so we make full use of all data and maximize based on four raters, with the largest number of observations statistical power for gene discovery. At the same time, by within age-bin around age 12 years. Within this age group, aggregating data at the level of the meta-analysis, we retain the phenotypic correlations among raters ranged between 0.22 and 1234567890();,: flexibility to estimate rgs between AGG at different ages, by 0.65, with a median of 0.34. The lowest phenotypic correlations different raters and instruments, and test how AGG assessed by were seen between teachers and parents. Since limited data multiple raters differ in the rg with other phenotypes. were available on individuals of non-European ancestry, we Data on AGG from parent-, teacher-, and self-report in boys and restricted analyses to individuals of European ancestry. girls were collected in 29 cohorts from Europe, USA, Australia, and In total, 29 cohorts contributed 163 GWASs, based on 328 935 New Zealand with 328 935 observations from 87 485 participants, observations from 87 485 unique individuals (Supplementary Table aged 1.5–18 years. First, we combined all data to produce the 2). Children were 1.5–18 years old at assessment, or retrospectively largest GWAMA on childhood AGG to date. SNP-based association assessed at these ages. Cohorts supplied between 1 and 26 tests were followed up by gene-based analyses. We computed univariate GWASs. Approximately 50% of the subjects were males. polygenic scores (PGSs) to test the out-of-sample prediction of Most GWASs were based on maternal- (52.4%) and self-assessment AGG to explore the usefulness of our GWAMA in future research (25.1%), with the remainder based on teacher (12.4%) and paternal [30]. To assess genetic pleiotropy between AGG and associated report (10.1%). After QC, applied to the univariate GWASs, between traits, we estimated rgs with a preselected set of external 3.47 M SNPs and 7.28 M SNPs were retained for meta-analysis (see phenotypes from multiple domains—with a focus on psychiatric Supplementary Fig. 2 and Supplementary Table 9). Note that the and psychological traits, cognition, anthropometric and reproduc- wide range of retained SNPs is a result of applying more stringent tive traits, substance use, and classic biomarkers of AGG, including QC filters for GWASs with smaller sample sizes and that GWASs testosterone levels. Second, meta-analyses were done by rater, with comparable sample sizes returned roughly equal number of instrument, and age. We estimated rgs across these assessments of SNPs (see Supplementary text and Supplementary Fig. 2). AGG. To identify context-specific genetic overlap with the external phenotypes, rgs were also estimated between rater-specific Meta-analysis assessments of AGG and the external phenotypes. Within cohort measures of AGG may be dependent due to including repeated measures of AGG over age and measures from multiple raters. To account for the effect of sample overlap, we METHODOLOGY applied a modified version of the multivariate meta-analysis Data description approach developed by Baselmans et al. [25] (see Table 1). Instead Extended description of the cohorts and phenotypes is supplied in of estimating the dependence among GWASs based on the cross- the Supplementary text and Supplementary Tables 1–9. Cohorts trait-intercept (CTI) with linkage disequilibrium score regression with assessment of AGG in genotyped children and adolescents (LDSC) [29, 31], the expected pairwise CTI value was calculated took part in the meta-analysis. AGG was assessed on continuous (Table 1) using the observed sample overlap and phenotypic scales, with higher scores indicating higher levels of AGG. Within covariance as sample sizes of the univariate GWASs were cohort, samples were stratified by (1) rater, (2) instrument, and (3) insufficient to run bivariate LDSC. The effective sample size (Neff) age, maintaining at least 450 observations in each stratum. We ran was approximated by the third formula in Table 1. When there is a univariate GWAS for each stratum within each cohort (Supple- no sample overlap (or a phenotypic correlation equal to zero) mentary Table 8). GWASs were run by local analysts following a between all GWASs (i.e., CTI is an identity matrix), Neff is equal to standard operation protocol (see URLs) after which the summary the sum of sample sizes. statistics were uploaded to a central location for the meta-analysis. First, we meta-analyzed all available GWASs (AGGoverall). Second, To account for dependence within cohort in the meta-analysis (see we meta-analyzed all available data within rater (rater-specific Supplementary text), each cohort supplied the phenotypic GWAMAs). Third, rater-specific age-bins were created for mother- covariance matrix between the AGG measures (Supplementary and self-reported AGG based on the mean ages of the subjects in Table 10) and the degree of sample overlap (Supplementary Table each GWAS (age-specific GWAMA). To ensure that the age-specific 11) between the different strata. Supplementary Fig. 1 shows the GWAMAs would have sufficient power for subsequent analyses, distribution of phenotypic correlations across all AGG measures. age-bins were created such that the total univariate number of We assumed no sample overlap across cohorts, and phenotypic observations (Nobs) exceeded 15 000 (see Supplementary text and Translational Psychiatry (2021)11:413 H.F. Ip et al. 3 Supplementary Table 12). For father- and teacher-reported AGG, relatedness estimated from SNPs and N = 10 706 is the number of there were insufficient data to run age-specific GWAMAs. Fourth, individuals. we performed instrument-specific GWAMAs for (1) the ASEBA scales and (2) for the SDQ, because for these two instruments the Genetic correlations with external phenotypes total univariate Nobs was over 15 000. We computed rgs between AGGoverall and a set of preselected SNPs that had MAF < 0.01, Neff < 15 000, or were observed in outcomes (N = 46; collectively referred to as “external pheno- less than two cohorts were removed from further analyses. SNP- types”; Supplementary Table 14). Phenotypes were selected based heritabilities (h2SNP ) were estimated using LDSC [31]. rgs were on established hypotheses with AGG and the availability of calculated across stratified assessments of AGG using LDSC [29]. sufficiently powered GWAS summary statistics. We restricted rgs to To ensure sufficient power for the genetic correlations, rg was phenotypes for which the Z-scores of the LDSC-based h2SNP ≥ 4 calculated across stratified assessments of AGG if the Z-score of [29]. Next, we estimated rgs for all rater-specific assessments of the h2SNP for the corresponding GWAMA was 4 or higher [29]. AGG (except for father-reported AGG). Genomic Structural Equation Modelling (Genomic SEM) [37] was applied to test if Gene-based tests rgs were significantly different across raters. Specifically, for every For AGGoverall, a gene-based analysis was done in MAGMA [32]. phenotype, we tested whether (1) all three rgs between the The gene-based test combines P values from multiple SNPs to external phenotype and rater-specific assessment of AGG, i.e., obtain a test statistic for each gene, while accounting for LD mother, teacher, or self-ratings, could be constrained at zero, and between the SNPs. From the MAGMA website (see URLs) we (2) whether rgs could be constrained to be equal across raters. A χ 2 obtained (1) a list of 18 087 genes and their start- and end- difference test was applied to assess whether imposing the positions, and (2) pre-formatted European genotypes from 1000 constraints resulted in a significant worse model fit compared to a Genomes phase 3 for the reference LD. We applied a Bonferroni model where the rgs between the phenotype and three rater- correction for multiple testing at α = 0.05/18 087. A lookup for specific assessment of AGG were allowed to differ. We applied an significant results was performed in GWAS Catalog and PhenoS- FDR correction at α = 0.05 over two models for 46 external canner (see URLs). phenotypes, for a total of 92 tests. An FDR correction for 4 × 46 = 184 tests was applied to correct for multiple testing of whether Polygenic scores the genetic correlations were significantly different from zero. All data were meta-analyzed twice more, once omitting all data from the Netherlands Twin Register (NTR) and once omitting the Australian data from the Queensland Institute for Medical RESULTS Research (QIMR) and the Mater-University of Queensland Study Overall GWAMA of Pregnancy (MUSP). As the NTR target sample we considered We first meta-analyzed the effect of each SNP across all available mother-reported AGG at age 7 (N = 4491), which represents the univariate GWASs. Assuming an Neff of 151 741, the h2SNP of largest NTR univariate stratum. In the QIRM participants, we tested AGGoverall was estimated at 3.31% (SE = 0.0038). The mean χ 2 whether our childhood AGG PGS predicted adult retrospective statistic was 1.12 along with an LDSC-intercept of 1.02 (SE = 0.01). assessment of their own CD behavior during adolescence (N = 10 This indicated that a small, but significant, part of the inflation in 706). We allowed for cohort-specific best practice in the PGS test statistics might have been due to confounding biases, which analysis. In the NTR, we created 16 sets of PGSs in PLINK1.9 [33], can either reflect population stratification or subtle misspecifica- with P value thresholds between 1 and 1.0E–05 (see Supplemen- tion of sample overlap within cohorts. No genome-wide tary Table 13). The remaining SNPs were clumped in PLINK. We significant hits were found for AGGoverall (Fig. 1). The list of applied an r2 threshold of 0.5 and minimum clumping distance of suggestive associations (P < 1.0E–05) is provided in Supplemen- 250 000 base pair positions [33]. Age, age [2], sex, first five tary Table 15. SNPs were annotated with SNPnexus (see URLs). The ancestry-based principal components, a SNP-array variable, and strongest association, in terms of significance, was located on interaction terms between sex and age, and sex and age [2] were chromosome 2 (rs2570485; P = 2.0E–07). The SNP is located inside defined as fixed effects. To account for relatedness, prediction was a gene desert, without any gene in 400 Kbp in any direction. The performed using generalized equation estimation (GEE) as second strongest independent association was found with implemented in the “gee” package (version 4.13–19) in R (version rs113599846 (P = 4.3E–07), which is located inside an intronic 3.5.3). GEE applies a sandwich correction over the standard errors to account for clustering in the data [34]. To correct for multiple testing, we applied an FDR correction at α = 0.05 for 16 tests. QIMR excluded SNPs with low imputation quality (r2 = 0.6) and MAF below 1% and selected the most significant independent SNPs using PLINK1.9 [35] (criteria linkage disequilibrium r2 = 0.1 within windows of 10 MBp). We calculated different PGS for seven P value thresholds (P < 1e–5, P < 0.001, P < 0.01, P < 0.05, P < 0.1, P < 0.5, and P < 1.0) of the GWAS summary statistics. PGS were calculated from the imputed genotype dosages to the 1000 Genomes (Phase 3 Release 5) reference panel. We fitted linear mixed models, which controlled for relatedness using a Genetic Relatedness Matrix (GRM) and covariates sex, age, two dummy variables for the GWAS array used, and the first five genetic principal components. The parameters of the model were estimated using GCTA 1.9 [36] The linear @@model was as follows: CDsymptom score ¼ intercept þ Covariates  b þ c  PGS þ G Fig. 1 Manhattan plot of overall meta-analysis for childhood aggression (AGGoverall). Red triangles represent SNPs that were Where b and c represent the vectors of fixed effects; and G  included in the significant genes from the gene-based analysis. SNPs Nð0; GRM  σ2GÞ represents the random effect that models the for ST3GAL3 and IPO13 are included in the same locus on sample relatedness, with GRM being the N by N matrix of chromosome 1. Translational Psychiatry (2021)11:413 H.F. Ip et al. 4 region of TNRC18 on chromosome 7. None of the suggestive associations have previously been reported for AGG or AGG- related traits [1]. We tested previously reported genome-wide significant asso- ciations for AGG [1] and performed a lookup in AGGoverall. We restricted lookup to associations with autosomal SNPs that were found in samples of European ancestry, resulting in three loci. One genome-wide significant hit was reported for adult antisocial personality disorder (rs4714329; OR = 0.63 [odds ratio was signed to the other allele in the original study]; P = 1.64E–09) [38]. The same SNP, however, had an opposite direction of effect in AGGoverall (β = 0.0022; P = 0.41). Tielbeek et al. [39] reported two genome-wide significant hits for antisocial behavior, one on chromosome 1 (rs2764450) and one on chromosome 11 (rs11215217). While both SNPs have the same direction of effect, neither SNP is associated with AGGoverall (both P > 0.5). Fig. 2 Proportion of explained variance (vertical axis) in child- Gene-based analysis hood aggression at age 7 by polygenic scores from the overall After correction for multiple testing, the gene-based analysis GWAMA for multiple P value thresholds (horizontal axis). returned three significant results (Supplementary Table 16): Numbers above the bars represent unadjusted P values for two- ST3GAL3 (ST3 beta-galactoside alpha-2,3-sialyltransferase3; P = sided test of significance. 1.6E–06), PCDH7 (protocadherin 7; P = 2.0E–06), and IPO13 (importin 13; P=2.5E–06). ST3GAL3 codes for a type II membrane protein that is involved in catalyzing the transfer of sialic acid from CMP-sialic acid to galactose-containing substrates. ST3GAL3 has been implicated in 107 GWASs, most notably on intelligence and educational attainment. The top SNP in ST3GAL3 (rs2485997; P = 2.48E–06) is in strong LD (r2 > 0.8) with several other SNPs inside the gene body of ST3GAL3 and in moderate LD (r2 > 0.6) with SNPs in several neighboring genes (Supplementary Fig. 3). PCDH7 codes for a protein that is hypothesized to function in cell-cell recognition and adhesion. PCDH7 has been implicated in 196 previous GWASs, for example educational attainment and adventurousness. The top SNP for PCDH7 (rs13138213; P = 1.44E–06) is in strong LD (r2 > 0.8) with a small number of other closely located SNPs and the signal for the gene-based test appears to be driven by two independent loci (Supplementary Fig. 4). IPO13 codes for a nuclear transport protein. IPO13 has been implicated in the UKB GWASs on whether a person holds a college or university degree and intelligence. The top SNP (rs3791116; P = 1.19E–05) is in moderate-to-strong LD with multiple SNPs (Supplementary Fig. 5), including SNPs in the neighboring ST3GAL3 gene. Polygenic prediction In children, 11 out of 16 PGSs were significantly correlated with mother-reported AGG in 7-year olds (Fig. 2) after correction for multiple testing. The scores explained between 0.036 and 0.44% of the phenotypic variance. The significant correlations consis- tently emerged when scores including SNPs with P values above 0.002 in the discovery GWAS were considered. In the retro- Fig. 3 Proportion of explained variance (vertical axis) in retro- spective assessments of adolescent CD, the PGS calculated at spective adolescent CD (two-sided tests). Blue bars indicate various thresholds (Fig. 3) explained up to 0.2% of the variance in positive correlation with the conduct disorder score. symptom sum scores. Generally, CD is significantly predicted at most thresholds, although, as we would expect based on the significant rg with AGGoverall (P < 0.02). In general, the highest SNP-heritability of AGGoverall, the proportion of explained positive correlations were seen with psychiatric traits, notably variance is small. ADHD, ASD, and major depressive disorder (MDD). The largest negative genetic correlations were found for age at smoking Genetic correlation with external phenotypes initiation, childhood IQ, and age at first birth. Based on the Genetic correlations between AGGoverall and a set of preselected biomarker-aggression literature, we tested for the presence of external phenotypes are shown in Fig. 4 and Supplementary Table genetic correlations between AGGoverall, and lipids, heart rate, 17. These phenotypes can broadly be grouped into psychiatric heart rate variability, and testosterone levels. Very low genetic and psychological traits, substance use, cognitive ability, anthro- correlations were observed for AGGoverall, and these biomarkers, pometric traits, classic biomarkers of AGG, reproductive traits, and with in many cases the sign of the genetic correlation opposite to sleeping behavior. We included childhood phenotypes (e.g., birth what was expected based on the literature on biomarkers of AGG. weight and childhood IQ) and disorders (e.g., ADHD and autism spectrum disorder [ASD]), but the majority of phenotypes were Stratified assessment of childhood aggressive behavior adult characteristics or characteristics measured in adult samples. Separate meta-analyses were carried out for raters, instruments After correction for multiple testing, 36 phenotypes showed a and age. None of these GWAMAs returned genome-wide Translational Psychiatry (2021)11:413 H.F. Ip et al. 5 Fig. 4 Genetic correlation with external phenotypes. Phenotypes are ordered by domain. Bars indicate 95% confidence intervals. significant hits. Manhattan plots for the four rater-specific Next, we tested for each external phenotype whether the three GWAMAs are shown in Supplementary Fig. 6. Estimates of h2SNP rater-specific rgs with the external phenotypes could be constrained for rater-specific assessment of AGG are shown in Supplementary to be equal across mothers, teachers and self-ratings. For ADHD, Table 18. The lowest h2SNP was observed for father-reported AGG ASD, MDD, schizophrenia, well-being, and self-reported health, (h2SNP = 0.04; SE = 0.03) and the highest for teacher-reported AGG constraining the rgs to be equal across rater resulted in significantly (h2SNP = 0.08; SE = 0.02). We estimated rg between rater-specific worse model fit (Supplementary Table 21). For all these phenotypes, assessment of AGG, except for father-reported AGG, which rgs with teacher-reported AGG were consistently lower compared to returned a non-significant h2SNP . A substantial genetic correlation mother- and self-reported AGG (Supplementary Fig. 7 and was observed between AGGMother and AGGTeacher (rg = 0.81; SE = Supplementary Table 17). For lifetime cannabis use, genetic 0.11). Moderate genetic correlations were observed between correlations also could not be constrained to be equal across AGGSelf and AGGMother (rg = 0.67; SE = 0.10), and between AGGSelf raters. Here, a relatively strong rg was found with self-reported AGG and AGGTeacher (rg = 0.46; SE = 0.13). Both genetic correlations (rg = 0.36; SE = 0.08) compared to teacher- (rg = 0.13; SE = 0.07) involving self-reported AGG were significantly lower than 1. and mother-reported AGG (rg = 0.08; SE = 0.08). We performed a GWAMA across all GWASs where an ASEBA scale was used (AGGASEBA) and another GWAMA across all GWASs for the SDQ (AGGSDQ). SNP-heritabilities for AGGASEBA and AGGSDQ DISCUSSION were 0.031 (SE = 0.0099) and 0.026 (SE = 0.0086), respectively. The We present the largest GWAMA of childhood AGG to date. The GWAMAs were insufficiently powered to estimate rg across gene-based analysis implicated three genes, PCDH7, ST3GAL3, and instrument-specific assessment of AGG. IPO13, based on the overall meta-analysis (AGGoverall), which did Age-specific GWAMAs were performed for mother- and self- not return genome-wide significant SNPs. Lead SNPs in the reported AGG, which made up 77.5% of the data. Mother- implicated genes were related to educational outcomes, but did reported data were split into seven age-bins and self-reported not reach genome-wide significance and these loci require further data into three (Supplementary Table 12). Estimates of the h2SNP evidence before being considered as AGG risk variants. PGS for each age-specific GWAMA can be found in Supplementary predicted childhood AGG and retrospectively assessed adolescent Table 19. For mother-reported AGG, h2SNP ranged between 0.012 CD. Stratified analyses within AGG generally returned moderate- and 0.078. For self-reported AGG, the highest h2SNP was seen for to-strong genetic correlations across raters. We found substantial the retrospective data (h2SNP = 0.12; SE = 0.03), which also genetic correlations between AGGoverall and a list of preselected showed a significantly inflated intercept (1.05; SE = 0.01). rg external phenotypes from various domains, including, psychiatry could only be estimated between AGGM7, AGGS13, and AGGSR and psychology, cognition, anthropometric and reproductive (Supplementary Table 20). traits. Most notably was the perfect rg between AGGoverall and ADHD (rg = 1.00; SE = 0.07). This is in line with the moderate-to- Genetic correlation between rater-specific assessment of AGG strong phenotypic correlations that have consistently been found and external phenotypes across sex-, rater-, age-, and instrument-specific assessment of We estimated rater-specific rgs with the external phenotypes, AGG with attention problems and hyperactivity [15]. Significant except for father-reported AGG, and tested for each external genetic correlations were further observed  with other psychiatric phenotype whether these rgs could be constrained to be equal to and psychological traits (range rg : 0.19–0.55). Negative genetic zero. For 31 out of 46 external phenotypes, constraining the rgs to correlations (rg = ~−0.5) were found with all three traits from the be equal to zero for all three raters resulted in significant cognitive domain. Genetic correlations were positive with smok- reduction in model fit (Supplementary Table 21), indicating that, ing initiation (rg = 0.55; SE = 0.04) and smoking quantity (rg = 0.46; for these external phenotypes, at least one rater has an rg that is SE = 0.06), and negative with age at smoking initiation (rg = significantly different from zero. −0.60; SE = 0.09). Translational Psychiatry (2021)11:413 H.F. Ip et al. 6 We examined genetic correlations with classical biomarkers of GWASs are increasingly successful in identifying genomic loci AGG. Higher levels of aggression have been associated with lower for complex human traits [48] and also in psychiatry, genetic levels of LDL [40] and lower resting heart rate [41, 42]. We found a biomarkers are increasingly thought of as promising for both positive, albeit weak, rg between AGGoverall and LDL (rg = 0.15; SE = research and treatment. Genetic risk prediction holds promise 0.07), which has an opposite sign than what was expected based on for adult psychiatric disorders [30] and it seems reasonable to the literature [39]. More broadly, except for HDL (rg = −0.13; SE = expect the same for childhood disorders. Here we found that 0.07), all measures of lipid levels returned significant positive rgs with PGSs explain up to 0.44% of the phenotypic variance in AGG in AGGoverall, albeit weakly (rg < 0.2). No heart rate measure showed a 7-year olds and 0.2% of the variance in retrospectively reported significant genetic correlation with AGGoverall. The relationship adolescent CD. Note that differences in ages, instrument and between testosterone levels and (childhood) AGG in the literature local best-practices have led to differences in explained is, at best, unclear. A positive association between AGG and variance. Future studies may explore the utility of these PGSs testosterone is often assumed, but the relation may be more in illuminating pleiotropy between AGGoverall and other traits. A complex [43]. Both positive and negative phenotypic correlations limiting factor in this regard is the relatively low SNP- have been found and seem context-dependent [44]. We found heritability, which puts an upper bound on the predictive significant negative, rgs between  AGGoverall and testosterone levels in accuracy of PGSs. Since measurement error suppresses SNP- males and females (rg  < 0.15). These should be interpreted with heritability, better measurement may offer an avenue to higher some caution because of the design of the GWA studies: AGG was powered GWAS, and subsequently to better PGS. Furthermore, measured in children and young adolescents whereas testosterone sample sizes for developmental phenotypes, including AGG, levels were measured in adults in the UK Biobank [45], and genetic may need to increase by one to two orders of magnitude before stability of testosterone levels might be low, at least for males [46]. PGS become useful for individual patients. Genetic correlations with reproductive traits showed a positive Despite our extensive effort, the first genome-wide significant relation with having more children (rg = 0.27; SE = 0.08) and having SNP for childhood AGG has yet to be found. Even in the absence of offspring earlier in life (rg = −0.60; SE = 0.06), tending to confirm that genome-wide significant loci, however, GWASs aid in clarifying the not all associated outcomes are harmful. biology behind complex traits. Our results show that, even without The stratified design of our study also allowed for examination genome-wide significant hits, a GWAS can be powerful enough to of the genetic etiology of AGG in subsets of the data and illuminate the genetic etiology of a trait in the form of rgs with other examination of genetic correlations among raters. We found a complex traits. Non-significant associations are expected to capture high genetic correlation between AGGMother and AGGTeacher (rg = part of the polygenicity of a trait [31] and various follow-up analyses 0.81; SE = 0.11). However, the 95% confidence interval covers 1, have been developed for GWASs that do not require, but are aided which makes these results hard to reconcile with previous findings by, genome-wide significant hits [49]. PGSs aggregate SNP effects of rater-specific additive genetic effects in childhood AGG [47]. into a weighted sum that indicates a person’s genetic liability to Most external phenotypes showed comparable rgs with mother-, develop a disorder. While their clinical application is still limited in self-, and teacher-reported AGG. For ADHD, ASD, MDD, schizo- psychiatric disorders, they can already aid in understanding the phrenia, well-being, and self-reported health, rgs differed sig- pleiotropy among psychiatric and other traits [30]. Similarly, nificantly across raters. Weaker rgs were consistently found in summary statistics-based genetic correlations (rg) provide insight teacher-reported AGG compared to mother- and self-reported into the genetic overlap between complex traits [29, 50]. AGG. These findings indicate the presence of rater-specific effects when considering the genetic correlation of AGG with other outcomes. rgs are generally stronger in the psychopathology and URLS psychological domains. A lack of power, however, seems GWAS SOP: http://www.action-euproject.eu/content/data-protocols insufficient to explain why we found weaker rgs between MAGMA: https://ctg.cncr.nl/software/magma AGGTeacher and phenotypes from these two domains. Other SNPnexus: https://www.snp-nexus.org/index.html (Accessed 28 phenotypes, like smoking behavior, educational attainment or Aug 2019) age at first birth, are, like psychopathological phenotypes, highly GWAS Catalog: https://www.ebi.ac.uk/gwas/ (Accessed 29 genetically correlated with AGGoverall, but, unlike psychopatholo- Aug 2019) gies, have near identical rgs across raters. The rater-specific effects PhenoScanner: http://www.phenoscanner.medschl.cam.ac.uk/ on rgs between childhood AGG and external phenotypes might be (Accessed 29 Aug 2019) limited to psychopathologies, and future research into the genetics of childhood psychopathology might consider these nuances in effects of assessment of childhood AGG from various CODE AVAILABILITY sources, be that multiple raters, instruments, and ages. Code for meta-analyses and follow-up analyses are available from the corresponding Despite the considerable sample sizes, we were still under- author upon request. powered to compute genetic correlations with external pheno- types while stratifying AGG over age or instrument. Age-stratified GWASs in larger samples across development are a desirable REFERENCES target for future research. Because genetic correlations can be 1. Odintsova VV, Roetman PJ, Ip HF, Pool R, Van der Laan CM, Tona KD, et al. computed between phenotypes for which a well-powered GWAS Genomics of human aggression: current state of genome-wide studies and an is available, age-stratified GWAS of many developmental pheno- automated systematic review tool. Psychiatr Genet. 2019;29:170–90. 2. Baron RA, Richardson DR. Human aggression. 2nd ed. Plenum Press: New York; types, behavioral, cognitive, and neuroscientific can be leveraged 1994. to better understand development of childhood traits. 3. Anderson CA, Bushman BJ. Human aggression. Annu Rev Psychol. 2002;53:27–51. We note that multivariate results should be interpreted with some 4. Veroude K, Zhang-James Y, Fernàndez-Castillo N, Bakker MJ, Cormand B, Faraone caution. While combining data from correlated traits can indeed SV. Genetics of aggressive behavior: an overview. Am J Med Genet Part B Neu- improve power to identify genome-wide associations, interpreting ropsychiatr Genet. 2016;171:3–43. the phenotype may not be straightforward. In the current GWAMA, 5. Tuvblad C, Baker LA. Human aggression across the lifespan: genetic propensities we have referred to our phenotype as “aggressive behavior” and and environmental moderators. Adv Genet. 2011;75:171–214. interpreted the results accordingly. AGG, however, is an umbrella 6. Bongers IL, Koot HM, van der Ende J, Verhulst FC. Developmental trajectories of term that has been used to identify a wide range of distinct—though externalizing behaviors in childhood and adolescence. Child Dev. 2004;75: 1523–37. correlated—traits and behaviors [1]. Translational Psychiatry (2021)11:413 H.F. Ip et al. 7 7. Tremblay RE, Vitaro F, Côté SM. Developmental origins of chronic physical 33. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation aggression: a bio-psycho-social model for the next generation of preventive PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4:7. interventions. Annu Rev Psychol. 2018;69:383–407. 34. Rogers P, Stoner J. Modification of the sandwich estimator in generalized esti- 8. Huesmann LR, Dubow EF, Boxer P. Continuity of aggression from childhood to mating equations with correlated binary outcomes in rare event and small early adulthood as a predictor of life outcomes: implications for the adolescent- sample settings. Am J Appl Math Stat. 2015;3:243–51. limited and life-course-persistent models. Aggress Behav. 2009;35:136–49. 35. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a 9. Zhang-James Y, Faraone SV. Genetic architecture for human aggression: a study tool set for whole-genome association and population-based linkage analyses. of gene-phenotype relationship in OMIM. Am J Med Genet Part B Neuropsychiatr Am J Hum Genet. 2007;81:559–75. Genet. 2016;171:641–9. 36. Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome-wide complex 10. American Psychiatric Association. Diagnostic and statistical manual of mental trait analysis. Am J Hum Genet. 2011;88:76–82.. disorders. 5th ed. Arlington: American Psychiatric Association; 2013. 37. Grotzinger AD, Rhemtulla M, de Vlaming R, Ritchie SJ, Mallard TT, Hill WD, et al. 11. Foster EM, Jones DE. The Conduct Problems Prevention Research Group. The Genomic structural equation modelling provides insights into the multivariate high costs of aggression: public expenditures resulting from conduct disorder. genetic architecture of complex traits. Nat Hum Behav. 2019;3:1–525. Am J Public Health. 2005;95:1767–72. 38. Rautiainen M-R, Paunio T, Repo-Tiihonen E, Virkkunen M, Ollila HM, Sulkava S, 12. Walton KE, Ormel J, Krueger RF. The dimensional nature of externalizing beha- et al. Genome-wide association study of antisocial personality disorder. Transl viors in adolescence: evidence from a direct comparison of categorical, dimen- Psychiatry. 2016;6:e883. sional, and hybrid models. J Abnorm Child Psychol. 2011;39:553–61. 39. Tielbeek JJ, Johansson A, Polderman T, Rautiainen MR, Jansen P, Taylor M, et al. 13. Walters GD, Ruscio J. Trajectories of youthful antisocial behavior: categories or Genome-wide association studies of a broad spectrum of antisocial behavior. continua? J. Abnorm Child Psychol. 2013;41:653–66. JAMA Psychiatry. 2017;74:1242–50. 14. Barry TD, Marcus DK, Barry CT, Coccaro EF. The latent structure of oppositional 40. Hagenbeek FA, et al. Adult aggressive behavior in humans and biomarkers: a defiant disorder in children andadults. J Psychiatr Res. 2013;47:1932–9. focus on lipids and methylation. J Pediatr Neonatal Individ Med. 2018;7:e070204. 15. Bartels M, Hendriks A, Mauri M, Krapohl E, Whipp A, Bolhuis K, et al. Childhood 41. Raine A, Fung ALC, Portnoy J, Choy O, Spring VL. Low heart rate as a risk factor for aggression and the co-occurrence of behavioural and emotional problems: child and adolescent proactive aggressive and impulsive psychopathic behavior. results across ages 3–16 years from multiple raters in six cohorts in the EU- Aggress Behav. 2014;40:290–9. ACTION project. Eur Child Adolesc Psychiatry. 2018;27:1105–21. 42. Latvala A, Kuja-Halkola R, Rück C, D'Onofrio BM, Jernberg T, Almqvist C, et al. 16. Whipp A, Vuoksimaa E, Bolhuis K, de Zeeuw E, Korhonen T, Mauri M, et al. Association of resting heart rate and blood pressure in late adolescence with Teacher-rated aggression and co-occurring problems and behaviors among subsequent mental disorders. JAMA Psychiatry. 2016;73:1268–75. schoolchildren: a comparison of four population-based European cohorts. 43. Brain PF, Susman EJ. Hormonal aspects of aggression and violence. In: Stoff, DM, medRxiv 2019:19002576. Breiling, J, Maser JD (Eds.). Handbook of antisocial behavior. John Wiley & Sons 17. Heron J, Barker ED, Joinson C, Lewis G, Hickman M, Munafò M, et al. Childhood Inc.: Hoboken; 1997;314–23. conduct disorder trajectories, prior risk factors and cannabis use at age 16: birth 44. Ramirez JM. Hormones and aggression in childhood and adolescence. Aggress cohort study. Addiction. 2013;108:2129–38. Violent Behav. 2003;8:621–44. 18. Rivenbark JG, Odgers CL, Caspi A, Harrington H, Hogan S, Houts RM, et al. The 45. Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank high societal costs of childhood conduct problems: evidence from administrative resource with deep phenotyping and genomic data. Nature. 2018;562:203–9. records up to age 38 in a longitudinal birth cohort. J Child Psychol Psychiatry. 46. Harris JA, Vernon PA, Boomsma DI. The heritability of testosterone: a study of 2018;59:703–10. Dutch adolescent twins and their parents. Behav Genet. 1998;28:165–71. 19. Odgers CL, Moffitt TE, Broadbent JM, Dickson N, Hancox RJ, Harrington H, et al. 47. Hudziak JJ, van Beijsterveldt CE, Bartels M, Rietveld MJ, Rettew DC, Derks EM, Female and male antisocial trajectories: from childhood origins to adult out- et al. Individual differences in aggression: genetic analyses by age, gender, and comes. Dev Psychopathol. 2008;20:673–716. informant in 3-, 7-, and 10-year-old Dutch twins. Behav Genet. 2003;33:575–89. 20. Whipp AM, Korhonen T, Raevuori A, Heikkilä K, Pulkkinen L, Rose RJ, et al. Early 48. Visscher PM, Wray NR, Zhang Q, Sklar P, McCarthy MI, Brown MA, et al. 10 years of adolescent aggression predicts antisocial personality disorder in young adults: a GWAS discovery: biology, function, and translation. Am J Hum Genet. population-based study. Eur Child Adolesc Psychiatry. 2019;28:341–50. 2017;101:5–22. 21. Hawley PH, Little TD, Rodkin PC. Aggression and adaptaion: the bright side to bad 49. Pasaniuc B, Price AL. Dissecting the genetics of complex traits using summary behavior. 1st ed. New York: Routledge; 2007 https://doi.org/10.4324/9780203936900 association statistics. Nat Rev Genet. 2017;18:117–27. 22. Hawley PH. Social dominance in childhood and adolescence: why social com- 50. Watanabe K, Stringer S, Frei O, Umićević Mirkov M, de Leeuw C, Polderman T, petence and aggression may go hand in hand. In: Hawley PH, Little TD, Rodkin et al. A global overview of pleiotropy and genetic architecture in complex traits. PC, editors. Aggression and adaptation: the bright side to bad behavior. New Nat Genet. 2019;51:1–10. York: Routledge; 2007. p. 1–22. 23. Waltes R, Chiocchetti AG, Freitag CM. The neurobiological basis of human aggression: a review on genetic and epigenetic mechanisms. Am J Med Genet ACKNOWLEDGEMENTS Part B Neuropsychiatr Genet. 2016;171:650–75. We very warmly thank all participants, their parents, and teachers for making this 24. Turley P, Walters RK, Maghzian O, Okbay A, Lee JJ, Fontana MA, et al. Multi-trait study possible. The project was supported by the “Aggression in Children: Unraveling analysis of genome-wide association summary statistics using MTAG. Nat Genet. gene-environment interplay to inform Treatment and InterventiON strategies” 2018;50:229–37. project (ACTION). ACTION received funding from the European Union Seventh 25. Baselmans BML, Jansen R, Ip HF, van Dongen J, Abdellaoui A, van de Weijer MP, Framework Program (FP7/2007-2013) under grant agreement no 602768. Cohort- et al. Multivariate genome-wide analyses of the well-being spectrum. Nat Genet. specific acknowledgements and funding information may be found in the 2019;51:1–451. Supplementary text. 26. van Beijsterveldt CEM, Bartels M, Hudziak JJ, Boomsma DI. Causes of stability of aggression from early childhood to adolescence: a longitudinal genetic analysis in Dutch twins. Behav Genet. 2003;33:591–605. 27. Porsch RM, Middeldorp CM, Cherny SS, Krapohl E, van Beijsterveldt CE, Loukola A, AUTHOR CONTRIBUTIONS et al. Longitudinal heritability of childhood aggression. Am J Med Genet Part B May be found in the Supplementary text. Neuropsychiatr Genet. 2016;171:697–707. 28. Lubke GH, McArtor DB, Boomsma DI, Bartels M. Genetic and environmental con- tributions to the development of childhood aggression. Dev Psychol. 2018;54:39–50. COMPETING INTERESTS 29. Bulik-Sullivan B, Finucane HK, Anttila V, Gusev A, Day FR, Loh PR, et al. An atlas of MC has received travel grants and research support from Eli Lilly and Co., Janssen- genetic correlations across human diseases and traits. Nat Genet. 2015;47:1236–41. Cilag, Shire, and Lundbeck and served as consultant for Eli Lilly and Co., Janssen- 30. Martin AR, Daly MJ, Robinson EB, Hyman SE, Neale BM. Predicting polygenic risk Cilag, Shire, and Lundbeck. JARQ was on the speakers’ bureau and/or acted as of psychiatric disorders. Biol Psychiatry. 2019;86:97–109. consultant Eli Lilly, Janssen-Cilag, Novartis, Shire, Lundbeck, Almirall, Braingaze, 31. Bulik-Sullivan BK, Loh PR, Finucane HK, Ripke S, Yang J. Schizophrenia Working Sincrolab, Medicine, Exeltis, and Rubió in the last 5 years. He also received travel Group of the Psychiatric Genomics Consortium, et al. LD Score regression dis- awards (air tickets + hotel) for taking part in psychiatric meetings from Janssen-Cilag, tinguishes confounding from polygenicity in genome-wide association studies. Rubió, Shire, Medice, and Eli-Lilly. The Department of Psychiatry chaired by him Nat Genet. 2015;47:291–95. received unrestricted educational and research support from the following 32. de Leeuw CA, Mooij JM, Heskes T, Posthuma D. MAGMA: generalized gene-set companies in the last 5 years: Eli-Lilly, Lundbeck, Janssen-Cilag, Actelion, Shire, analysis of GWAS data. PLOS Comput Biol. 2015;11:e1004219. Ferrer, Oryzon, Roche, Psious, and Rubió. Translational Psychiatry (2021)11:413 H.F. Ip et al. 8 ADDITIONAL INFORMATION Open Access This article is licensed under a Creative Commons Supplementary information The online version contains supplementary material Attribution 4.0 International License, which permits use, sharing, available at https://doi.org/10.1038/s41398-021-01480-x. adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Correspondence and requests for materials should be addressed to M.B., M.G.N. or Commons license, and indicate if changes were made. The images or other third party D.I.B. material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the Reprints and permission information is available at http://www.nature.com/ article’s Creative Commons license and your intended use is not permitted by statutory reprints regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims org/licenses/by/4.0/. in published maps and institutional affiliations. © The Author(s) 2021 Hill F. Ip 1, Camiel M. van der Laan1,2, Eva M. L. Krapohl3, Isabell Brikell4, Cristina Sánchez-Mora 5,6,7, Ilja M. Nolte 8, Beate St Pourcain 9,10,11, Koen Bolhuis12, Teemu Palviainen 13, Hadi Zafarmand14,15, Lucía Colodro-Conde 16, Scott Gordon 16, Tetyana Zayats17,18,19, Fazil Aliev 20,21, Chang Jiang22,23, Carol A. Wang24, Gretchen Saunders25, Ville Karhunen 26, Anke R. Hammerschlag1,27,28, Daniel E. Adkins29,30, Richard Border 31,32,33, Roseann E. Peterson34, Joseph A. Prinz35, Elisabeth Thiering36,37, Ilkka Seppälä38, Natàlia Vilor-Tejedor39,40,41,42,43, Tarunveer S. Ahluwalia 44,45, Felix R. Day 46, Jouke-Jan Hottenga 1, Andrea G. Allegrini3, Kaili Rimfeld 3, Qi Chen4, Yi Lu 4, Joanna Martin4,47, María Soler Artigas 5,6,7, Paula Rovira5,6,7, Rosa Bosch 5,6,48, Gemma Español5, Josep Antoni Ramos Quiroga 5,6,7,48, Alexander Neumann 12,49, Judith Ensink50,51, Katrina Grasby 16, José J. Morosoli16, Xiaoran Tong22,23, Shelby Marrington52, Christel Middeldorp 1,27,53, James G. Scott52,54,55, Anna Vinkhuyzen56, Andrey A. Shabalin30, Robin Corley31,57, Luke M. Evans 31,57, Karen Sugden35,58, Silvia Alemany 39,40,41, Lærke Sass 44, Rebecca Vinding44, Kate Ruth 59, Jess Tyrrell59, Gareth E. Davies60, Erik A. Ehli 60, Fiona A. Hagenbeek 1, Eveline De Zeeuw1, Toos C.E.M. Van Beijsterveldt1, Henrik Larsson4,61, Harold Snieder 8, Frank C. Verhulst12,62, Najaf Amin63, Alyce M. Whipp13, Tellervo Korhonen13, Eero Vuoksimaa13, Richard J. Rose64, André G. Uitterlinden 63,65,66, Andrew C. Heath67, Pamela Madden67, Jan Haavik 17,68, Jennifer R. Harris69, Øyvind Helgeland70, Stefan Johansson 17,71, Gun Peggy S. Knudsen 69, Pal Rasmus Njolstad 72, Qing Lu22,23, Alina Rodriguez 26,73, Anjali K. Henders56, Abdullah Mamun74, Jackob M. Najman52, Sandy Brown 75, Christian Hopfer76, Kenneth Krauter77, Chandra Reynolds78, Andrew Smolen31, Michael Stallings31,32, Sally Wadsworth31, Tamara L. Wall75, Judy L. Silberg34,79, Allison Miller 80, Liisa Keltikangas-Järvinen 81 , Christian Hakulinen81, Laura Pulkki-Råback81, Alexandra Havdahl 82,83, Per Magnus84, Olli T. Raitakari85,86,87, John R. B. Perry 46, Sabrina Llop88,89, Maria-Jose Lopez-Espinosa88,89,90, Klaus Bønnelykke44, Hans Bisgaard44, Jordi Sunyer39,40,41,91, Terho Lehtimäki 38, Louise Arseneault 92, Marie Standl36, Joachim Heinrich36,93,94, Joseph Boden 95, John Pearson 96, L. John Horwood95, Martin Kennedy 97, Richie Poulton 98, Lindon J. Eaves34,79, Hermine H. Maes34,79,99, John Hewitt31,32, William E. Copeland 100, Elizabeth J. Costello101, Gail M. Williams52, Naomi Wray 56,102, Marjo-Riitta Järvelin 26,103, Matt McGue 25, William Iacono25, Avshalom Caspi35,58,92,104, Terrie E. Moffitt35,58,92,104, Andrew Whitehouse105, Craig E. Pennell 24, Kelly L. Klump106, S. Alexandra Burt106, Danielle M. Dick 20,107,108, Ted Reichborn-Kjennerud83,109, Nicholas G. Martin 16, Sarah E. Medland 16, Tanja Vrijkotte14, Jaakko Kaprio 13,110, Henning Tiemeier 12,111, George Davey Smith 9,112, Catharina A. Hartman113, Albertine J. Oldehinkel 113, Miquel Casas5,6,7,48, Marta Ribasés 5,6,7, Paul Lichtenstein 4, Sebastian Lundström114,115, Robert Plomin 3, Meike Bartels 1,28,231 ✉, Michel G. Nivard 1,231 ✉ and Dorret I. Boomsma 1,28,231 ✉ 1 Department of Biological Psychology, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands. 2Netherlands Institute for the Study of Crime and Law Enforcement, Amsterdam, The Netherlands. 3Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, UK. 4Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden. 5Department of Psychiatry, Hospital Universitari Vall d’Hebron, Barcelona, Spain. 6Biomedical Network Research Centre on Mental Health (CIBERSAM), Instituto de Salud Carlos III, Barcelona, Spain. 7Psychiatric Genetics Unit, Group of Psychiatry, Mental Health and Addiction, Vall d’Hebron Research Institute (VHIR), Universitat Autònoma de Barcelona, Barcelona, Spain. 8Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands. 9MRC Integrative Epidemiology Unit, University of Bristol, Bristol, UK. 10Max Planck Institute for Psycholinguistics, Nijmegen, The Netherlands. 11Donders Institute for Brain, Cognition and Behaviour, Radboud University, Nijmegen, The Netherlands. 12Department of Child and Adolescent Psychiatry/Psychology, Erasmus University Medical Center, Rotterdam, The Netherlands. 13Institute for Molecular Medicine FIMM, HiLife, University of Helsinki, Helsinki, Finland. 14Department of Public Health, Amsterdam Public Health Research Institute, Amsterdam UMC, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands. 15Department of Clinical Epidemiology, Biostatistics and Bioinformatics, Amsterdam Public Health Research Institute, Amsterdam UMC, Academic Medical Center, University of Amsterdam, Amsterdam, The Netherlands. 16 QIMR Berghofer Medical Research Institute, Brisbane, QLD, Australia. 17Department of Biomedicine, University of Bergen, Bergen, Norway. 18Analytic and Translational Genetics Unit, Department of Medicine, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA. 19Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, Cambridge, MA, USA. 20Department of Psychology, Virginia Commonwealth University, Richmond, VA, USA. 21Faculty of Business, Karabuk University, Karabuk, Turkey. 22 Department of Epidemiology and Biostatistics, Michigan State University, East Lansing, MI, USA. 23Department of Biostatistics, University of Florida, Gainesville, FL, USA. 24 Faculty of Medicine and Health, School of Medicine and Public Health, The University of Newcastle, Callaghan, NSW, Australia. 25Department of Psychology, University of Minnesota, Minneapolis, MN, USA. 26Department of Epidemiology and Biostatistics, MRC-PHE Centre for Environment and Health, School of Public Health, Imperial College London, London, UK. 27Child Health Research Centre, the University of Queensland, Brisbane, QLD, Australia. 28Amsterdam Public Health Research Institute, Amsterdam, The Netherlands. 29Department of Sociology, College of Social and Behavioral Science, University of Utah, Salt Lake City, UT, USA. 30Department of Psychiatry, School of Medicine, University of Utah, Salt Lake City, UT, USA. 31Institute for Behavioral Genetics, University of Colorado Boulder, Boulder, CO, USA. 32Department of Psychology and Neuroscience, University of Colorado Boulder, Boulder, CO, USA. 33Department of Applied Mathematics, University of Colorado Boulder, Boulder, CO, USA. 34Department of Psychiatry, Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Richmond, VA, USA. 35Center for Genomic and Computational Biology, Duke University, Durham, NC, USA. 36Institute of Epidemiology, Helmholtz Zentrum München – German Research Center for Environmental Health, Neuherberg, Germany. 37Division of Metabolic Diseases and Nutritional Medicine, Dr. von Hauner Children’s Hospital, Ludwig-Maximilians-University of Munich, Munich, Germany. 38Department of Clinical Chemistry, Fimlab Laboratories, and Finnish Cardiovascular Research Center – Tampere, Faculty of Medicine and Health Technology, Tampere University, Tampere, Finland. 39ISGlobal, Barcelona Institute for Global Health, Barcelona, Spain. 40Universitat Pompeu Fabra (UPF), Barcelona, Spain. 41CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain. 42Centre for Genomic Regulation (CRG), The Barcelona Institute of Science and Technology, Barcelona, Spain. 43Barcelona Beta Brain Research Center, Pasqual Maragall Foundation (FPM), Translational Psychiatry (2021)11:413 H.F. Ip et al. 9 Barcelona, Spain. 44COPSAC, Copenhagen Prospective Studies on Asthma in Childhood, Herlev and Gentofte Hospital, University of Copenhagen, Copenhagen, Denmark. 45Steno Diabetes Center Copenhagen, Gentofte, Denmark. 46MRC Epidemiology Unit, Institute of Metabolic Science, Cambridge Biomedical Campus, University of Cambridge School of Clinical Medicine, Cambridge, UK. 47MRC Centre for Neuropsychiatric Genetics and Genomics, Division of Psychological Medicine and Clinical Neurosciences, Cardiff University, Cardiff, UK. 48Department of Psychiatry and Legal Medicine, Universitat Autònoma de Barcelona, Barcelona, Spain. 49Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, QC, Canada. 50Department of Child and Adolescent Psychiatry, Academic Medical Center, Amsterdam, The Netherlands. 51De Bascule, Academic Centre for Child and Adolescent Psychiatry, Amsterdam, The Netherlands. 52School of Public Health, Faculty of Medicine, The University of Queensland, Herston, QLD, Australia. 53Children’s Health Queensland Hospital and Health Service, Child and Youth Mental Health Service, Brisbane, QLD, Australia. 54Metro North Mental Health, University of Queensland, St Lucia, QLD, Australia. 55Queensland Centre for Mental Health Research, St Lucia, QLD, Australia. 56Institute for Molecular Bioscience, University of Queensland, St Lucia, QLD, Australia. 57 Department of Ecology and Evolutionary Biology, University of Colorado Boulder, Boulder, CO, USA. 58Department of Psychology and Neuroscience and Center for Genomic and Computational Biology, Duke University, Durham, NC, USA. 59Genetics of Complex Traits, Royal Devon & Exeter Hospital, University of Exeter Medical School, Exeter, UK. 60 Avera Institute for Human Genetics, Sioux Falls, SD, USA. 61School of Medical Sciences, Orebro University, Orebro, Sweden. 62Child and Adolescent Mental Health Centre, Mental Health Services Capital Region, Research Unit, Copenhagen University Hospital, Copenhagen, Denmark. 63Department of Epidemiology, Erasmus University Medical Center, Rotterdam, The Netherlands. 64Department of Psychological and Brain Sciences, Indiana University, Bloomington, IN, USA. 65Department of Internal Medicine, Erasmus University Medical Center, Rotterdam, The Netherlands. 66Netherlands Genomics Initiative (NGI)-sponsored Netherlands Consortium for Healthy Aging (NCHA), Leiden, The Netherlands. 67 Washington University, St Louis, MO, USA. 68Division of Psychiatry, Haukeland University Hospital, Bergen, Norway. 69Division of Health Data and Digitalisation, The Norwegian Institute of Public Health, Oslo, Norway. 70Department of Genetics and Bioinformatics, Division of Health Data and Digitalization, The Norwegian Institute of Public Health, Bergen, Norway. 71K.G. Jebsen Centre for Neuropsychiatric Disorders, Department of Clinical Science, University of Bergen, Bergen, Norway. 72Department of Clinical Science, University of Bergen, Bergen, Norway. 73School of Psychology, University of Lincoln, Lincolnshire, UK. 74Institute for Social Science Research, University of Queensland, Long Pocket, QLD, Australia. 75Department of Psychiatry, University of California, San Diego, CA, USA. 76University of Colorado School of Medicine, Aurora, CO, USA. 77Department of Molecular, Cellular, and Developmental Biology, University of Colorado Boulder, Boulder, CO, USA. 78Department of Psychology, University of California Riverside, Riverside, CA, USA. 79 Department of Human & Molecular Genetics, Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Richmond, VA, USA. 80Department of Pathology and Biomedical Science, and Carney Centre for Pharmacogenomics, University of Otago Christchurch, Christchurch Central City, New Zealand. 81Department of Psychology and Logopedics, Faculty of Medicine, University of Helsinki, Helsinki, Finland. 82Nic Waals Institute, Lovisenberg Diaconal Hospital, Oslo, Norway. 83Department of Mental Disorders, Norwegian Institute of Public Health, Oslo, Norway. 84Centre for Fertility and Health, Norwegian Institute of Public Health, Oslo, Norway. 85Centre for Population Health Research, University of Turku and Turku University Hospital, Turku, Finland. 86Research Centre of Applied and Preventive Cardiovascular Medicine, University of Turku, Turku, Finland. 87Department of Clinical Physiology and Nuclear Medicine, Turku University Hospital, Turku, Finland. 88Epidemiology and Environmental Health Joint Research Unit, FISABIO-Universitat Jaume I-Universitat de València, Valencia, Spain. 89Spanish Consortium for Research on Epidemiology and Public Health (CIBERESP), Madrid, Spain. 90 Faculty of Nursing and Chiropody, Universitat de València, Valencia, Spain. 91IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain. 92Department of Psychiatry and Behavioral Sciences, Duke University School of Medicine, Durham, NC, USA. 93Institute and Outpatient Clinic for Occupational, Social and Environmental Medicine, University of Munich Medical Center, Ludwig-Maximilians-Universität München, Munich, Germany. 94Allergy and Lung Health Unit, Melbourne School of Population and Global Health, University of Melbourne, Melbourne, VIC, Australia. 95Christchurch Health and Development Study, Department of Psychological Medicine, University of Otago Christchurch, Christchurch Central City, New Zealand. 96Biostatistics and Computational Biology Unit, Department of Pathology and Biomedical Science, University of Otago Christchurch, Christchurch Central City, New Zealand. 97Department of Pathology and Biomedical Science, and Carney Centre for Pharmacogenomics, University of Otago Christchurch, Christchurch Central City, New Zealand. 98Dunedin Multidisciplinary Health and Development Research Unit, University of Otago, Dunedin, New Zealand. 99Massey Cancer Center, Virginia Commonwealth University, Richmond, VA, USA. 100Department of Psychiatry, College of Medicine, University of Vermont, Burlington, VT, USA. 101Department of Psychiatry, School of Medicine, Duke University, Durham, UK. 102Queensland Brain Institute, Institute for Molecular Bioscience, University of Queensland, St Lucia, QLD, Australia. 103 Center for Life Course Health Research, Faculty of Medicine, University of Oulu, Oulu, Finland. 104Social, Genetic, and Developmental Psychiatry Research Centre, Institute of Psychiatry, Psychology, and Neuroscience, King’s College London, London, UK. 105Telethon Kids Institute, The University of Western Australia, Perth, WA, Australia. 106Department of Psychology, Michigan State University, East Lansing, MI, USA. 107Department of Human and Molecular Genetics, Virginia Commonwealth University, Richmond, VA, USA. 108 College Behavioral and Emotional Health Institute, Virginia Commonwealth University, Richmond, VA, USA. 109Institute of Clinical Medicine, University of Oslo, Oslo, Norway. 110 Department of Public Health, Medical Faculty, University of Helsinki, Helsinki, Finland. 111Department of Social and Behavioral Science, Harvard TH Chan School of Public Health, Boston, MA, USA. 112Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, UK. 113Department of Psychiatry, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands. 114Gillberg Neuropsychiatry Centre, University of Gothenburg, Gothenburg, Sweden. 115Centre for Ethics, Law and Mental Health, University of Gothenburg, Gothenburg, Sweden. 116These authors jointly supervised this work: Meike Bartels, Michel G. Nivard, Dorret I. Boomsma. ✉email: m. bartels@vu.nl; m.g.nivard@vu.nl; di.boomsma@vu.nl Translational Psychiatry (2021)11:413